ffmpeg/libavcodec/rv40dsp.c
Kostya Shishkov a80852316e RV40: reuse some H.264 motion compensation functions in RV40.
Since RV40 is based on H.264, it uses the same interpolation filters and
functions for halfpel interpolation in one direction can be reused.
Unfortunately other cases are implemented differently and produce slightly
different output because of rounding.

Signed-off-by: Mans Rullgard <mans@mansr.com>
2011-08-30 12:04:46 +01:00

374 lines
17 KiB
C

/*
* RV40 decoder motion compensation functions
* Copyright (c) 2008 Konstantin Shishkov
*
* This file is part of Libav.
*
* Libav is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* Libav is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with Libav; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
/**
* @file
* RV40 decoder motion compensation functions
*/
#include "avcodec.h"
#include "dsputil.h"
#include "rv34dsp.h"
#define RV40_LOWPASS(OPNAME, OP) \
static av_unused void OPNAME ## rv40_qpel8_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
const int h, const int C1, const int C2, const int SHIFT){\
uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
int i;\
for(i = 0; i < h; i++)\
{\
OP(dst[0], (src[-2] + src[ 3] - 5*(src[-1]+src[2]) + src[0]*C1 + src[1]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[1], (src[-1] + src[ 4] - 5*(src[ 0]+src[3]) + src[1]*C1 + src[2]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[2], (src[ 0] + src[ 5] - 5*(src[ 1]+src[4]) + src[2]*C1 + src[3]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[3], (src[ 1] + src[ 6] - 5*(src[ 2]+src[5]) + src[3]*C1 + src[4]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[4], (src[ 2] + src[ 7] - 5*(src[ 3]+src[6]) + src[4]*C1 + src[5]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[5], (src[ 3] + src[ 8] - 5*(src[ 4]+src[7]) + src[5]*C1 + src[6]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[6], (src[ 4] + src[ 9] - 5*(src[ 5]+src[8]) + src[6]*C1 + src[7]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[7], (src[ 5] + src[10] - 5*(src[ 6]+src[9]) + src[7]*C1 + src[8]*C2 + (1<<(SHIFT-1))) >> SHIFT);\
dst += dstStride;\
src += srcStride;\
}\
}\
\
static void OPNAME ## rv40_qpel8_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
const int w, const int C1, const int C2, const int SHIFT){\
uint8_t *cm = ff_cropTbl + MAX_NEG_CROP;\
int i;\
for(i = 0; i < w; i++)\
{\
const int srcB = src[-2*srcStride];\
const int srcA = src[-1*srcStride];\
const int src0 = src[0 *srcStride];\
const int src1 = src[1 *srcStride];\
const int src2 = src[2 *srcStride];\
const int src3 = src[3 *srcStride];\
const int src4 = src[4 *srcStride];\
const int src5 = src[5 *srcStride];\
const int src6 = src[6 *srcStride];\
const int src7 = src[7 *srcStride];\
const int src8 = src[8 *srcStride];\
const int src9 = src[9 *srcStride];\
const int src10 = src[10*srcStride];\
OP(dst[0*dstStride], (srcB + src3 - 5*(srcA+src2) + src0*C1 + src1*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[1*dstStride], (srcA + src4 - 5*(src0+src3) + src1*C1 + src2*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[2*dstStride], (src0 + src5 - 5*(src1+src4) + src2*C1 + src3*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[3*dstStride], (src1 + src6 - 5*(src2+src5) + src3*C1 + src4*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[4*dstStride], (src2 + src7 - 5*(src3+src6) + src4*C1 + src5*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[5*dstStride], (src3 + src8 - 5*(src4+src7) + src5*C1 + src6*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[6*dstStride], (src4 + src9 - 5*(src5+src8) + src6*C1 + src7*C2 + (1<<(SHIFT-1))) >> SHIFT);\
OP(dst[7*dstStride], (src5 + src10 - 5*(src6+src9) + src7*C1 + src8*C2 + (1<<(SHIFT-1))) >> SHIFT);\
dst++;\
src++;\
}\
}\
\
static void OPNAME ## rv40_qpel16_v_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
const int w, const int C1, const int C2, const int SHIFT){\
OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\
OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
src += 8*srcStride;\
dst += 8*dstStride;\
OPNAME ## rv40_qpel8_v_lowpass(dst , src , dstStride, srcStride, w-8, C1, C2, SHIFT);\
OPNAME ## rv40_qpel8_v_lowpass(dst+8, src+8, dstStride, srcStride, w-8, C1, C2, SHIFT);\
}\
\
static void OPNAME ## rv40_qpel16_h_lowpass(uint8_t *dst, uint8_t *src, int dstStride, int srcStride,\
const int h, const int C1, const int C2, const int SHIFT){\
OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, 8, C1, C2, SHIFT);\
OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, 8, C1, C2, SHIFT);\
src += 8*srcStride;\
dst += 8*dstStride;\
OPNAME ## rv40_qpel8_h_lowpass(dst , src , dstStride, srcStride, h-8, C1, C2, SHIFT);\
OPNAME ## rv40_qpel8_h_lowpass(dst+8, src+8, dstStride, srcStride, h-8, C1, C2, SHIFT);\
}\
\
#define RV40_MC(OPNAME, SIZE) \
static void OPNAME ## rv40_qpel ## SIZE ## _mc10_c(uint8_t *dst, uint8_t *src, int stride){\
OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc30_c(uint8_t *dst, uint8_t *src, int stride){\
OPNAME ## rv40_qpel ## SIZE ## _h_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc01_c(uint8_t *dst, uint8_t *src, int stride){\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 52, 20, 6);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc11_c(uint8_t *dst, uint8_t *src, int stride){\
uint8_t full[SIZE*(SIZE+5)];\
uint8_t * const full_mid = full + SIZE*2;\
put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc21_c(uint8_t *dst, uint8_t *src, int stride){\
uint8_t full[SIZE*(SIZE+5)];\
uint8_t * const full_mid = full + SIZE*2;\
put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc31_c(uint8_t *dst, uint8_t *src, int stride){\
uint8_t full[SIZE*(SIZE+5)];\
uint8_t * const full_mid = full + SIZE*2;\
put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 52, 20, 6);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc12_c(uint8_t *dst, uint8_t *src, int stride){\
uint8_t full[SIZE*(SIZE+5)];\
uint8_t * const full_mid = full + SIZE*2;\
put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc22_c(uint8_t *dst, uint8_t *src, int stride){\
uint8_t full[SIZE*(SIZE+5)];\
uint8_t * const full_mid = full + SIZE*2;\
put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc32_c(uint8_t *dst, uint8_t *src, int stride){\
uint8_t full[SIZE*(SIZE+5)];\
uint8_t * const full_mid = full + SIZE*2;\
put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 52, 6);\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 20, 5);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc03_c(uint8_t *dst, uint8_t *src, int stride){\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, src, stride, stride, SIZE, 20, 52, 6);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc13_c(uint8_t *dst, uint8_t *src, int stride){\
uint8_t full[SIZE*(SIZE+5)];\
uint8_t * const full_mid = full + SIZE*2;\
put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 52, 20, 6);\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
}\
\
static void OPNAME ## rv40_qpel ## SIZE ## _mc23_c(uint8_t *dst, uint8_t *src, int stride){\
uint8_t full[SIZE*(SIZE+5)];\
uint8_t * const full_mid = full + SIZE*2;\
put_rv40_qpel ## SIZE ## _h_lowpass(full, src - 2*stride, SIZE, stride, SIZE+5, 20, 20, 5);\
OPNAME ## rv40_qpel ## SIZE ## _v_lowpass(dst, full_mid, stride, SIZE, SIZE, 20, 52, 6);\
}\
\
#define op_avg(a, b) a = (((a)+cm[b]+1)>>1)
#define op_put(a, b) a = cm[b]
RV40_LOWPASS(put_ , op_put)
RV40_LOWPASS(avg_ , op_avg)
#undef op_avg
#undef op_put
RV40_MC(put_, 8)
RV40_MC(put_, 16)
RV40_MC(avg_, 8)
RV40_MC(avg_, 16)
static const int rv40_bias[4][4] = {
{ 0, 16, 32, 16 },
{ 32, 28, 32, 28 },
{ 0, 32, 16, 32 },
{ 32, 28, 32, 28 }
};
#define RV40_CHROMA_MC(OPNAME, OP)\
static void OPNAME ## rv40_chroma_mc4_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
const int A = (8-x) * (8-y);\
const int B = ( x) * (8-y);\
const int C = (8-x) * ( y);\
const int D = ( x) * ( y);\
int i;\
int bias = rv40_bias[y>>1][x>>1];\
\
assert(x<8 && y<8 && x>=0 && y>=0);\
\
if(D){\
for(i = 0; i < h; i++){\
OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
dst += stride;\
src += stride;\
}\
}else{\
const int E = B + C;\
const int step = C ? stride : 1;\
for(i = 0; i < h; i++){\
OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
dst += stride;\
src += stride;\
}\
}\
}\
\
static void OPNAME ## rv40_chroma_mc8_c(uint8_t *dst/*align 8*/, uint8_t *src/*align 1*/, int stride, int h, int x, int y){\
const int A = (8-x) * (8-y);\
const int B = ( x) * (8-y);\
const int C = (8-x) * ( y);\
const int D = ( x) * ( y);\
int i;\
int bias = rv40_bias[y>>1][x>>1];\
\
assert(x<8 && y<8 && x>=0 && y>=0);\
\
if(D){\
for(i = 0; i < h; i++){\
OP(dst[0], (A*src[0] + B*src[1] + C*src[stride+0] + D*src[stride+1] + bias));\
OP(dst[1], (A*src[1] + B*src[2] + C*src[stride+1] + D*src[stride+2] + bias));\
OP(dst[2], (A*src[2] + B*src[3] + C*src[stride+2] + D*src[stride+3] + bias));\
OP(dst[3], (A*src[3] + B*src[4] + C*src[stride+3] + D*src[stride+4] + bias));\
OP(dst[4], (A*src[4] + B*src[5] + C*src[stride+4] + D*src[stride+5] + bias));\
OP(dst[5], (A*src[5] + B*src[6] + C*src[stride+5] + D*src[stride+6] + bias));\
OP(dst[6], (A*src[6] + B*src[7] + C*src[stride+6] + D*src[stride+7] + bias));\
OP(dst[7], (A*src[7] + B*src[8] + C*src[stride+7] + D*src[stride+8] + bias));\
dst += stride;\
src += stride;\
}\
}else{\
const int E = B + C;\
const int step = C ? stride : 1;\
for(i = 0; i < h; i++){\
OP(dst[0], (A*src[0] + E*src[step+0] + bias));\
OP(dst[1], (A*src[1] + E*src[step+1] + bias));\
OP(dst[2], (A*src[2] + E*src[step+2] + bias));\
OP(dst[3], (A*src[3] + E*src[step+3] + bias));\
OP(dst[4], (A*src[4] + E*src[step+4] + bias));\
OP(dst[5], (A*src[5] + E*src[step+5] + bias));\
OP(dst[6], (A*src[6] + E*src[step+6] + bias));\
OP(dst[7], (A*src[7] + E*src[step+7] + bias));\
dst += stride;\
src += stride;\
}\
}\
}
#define op_avg(a, b) a = (((a)+((b)>>6)+1)>>1)
#define op_put(a, b) a = ((b)>>6)
RV40_CHROMA_MC(put_, op_put)
RV40_CHROMA_MC(avg_, op_avg)
#define RV40_WEIGHT_FUNC(size) \
static void rv40_weight_func_ ## size (uint8_t *dst, uint8_t *src1, uint8_t *src2, int w1, int w2, int stride)\
{\
int i, j;\
\
for (j = 0; j < size; j++) {\
for (i = 0; i < size; i++)\
dst[i] = (((w2 * src1[i]) >> 9) + ((w1 * src2[i]) >> 9) + 0x10) >> 5;\
src1 += stride;\
src2 += stride;\
dst += stride;\
}\
}
RV40_WEIGHT_FUNC(16)
RV40_WEIGHT_FUNC(8)
av_cold void ff_rv40dsp_init(RV34DSPContext *c, DSPContext* dsp) {
c->put_pixels_tab[0][ 0] = dsp->put_h264_qpel_pixels_tab[0][0];
c->put_pixels_tab[0][ 1] = put_rv40_qpel16_mc10_c;
c->put_pixels_tab[0][ 2] = dsp->put_h264_qpel_pixels_tab[0][2];
c->put_pixels_tab[0][ 3] = put_rv40_qpel16_mc30_c;
c->put_pixels_tab[0][ 4] = put_rv40_qpel16_mc01_c;
c->put_pixels_tab[0][ 5] = put_rv40_qpel16_mc11_c;
c->put_pixels_tab[0][ 6] = put_rv40_qpel16_mc21_c;
c->put_pixels_tab[0][ 7] = put_rv40_qpel16_mc31_c;
c->put_pixels_tab[0][ 8] = dsp->put_h264_qpel_pixels_tab[0][8];
c->put_pixels_tab[0][ 9] = put_rv40_qpel16_mc12_c;
c->put_pixels_tab[0][10] = put_rv40_qpel16_mc22_c;
c->put_pixels_tab[0][11] = put_rv40_qpel16_mc32_c;
c->put_pixels_tab[0][12] = put_rv40_qpel16_mc03_c;
c->put_pixels_tab[0][13] = put_rv40_qpel16_mc13_c;
c->put_pixels_tab[0][14] = put_rv40_qpel16_mc23_c;
c->put_pixels_tab[0][15] = ff_put_rv40_qpel16_mc33_c;
c->avg_pixels_tab[0][ 0] = dsp->avg_h264_qpel_pixels_tab[0][0];
c->avg_pixels_tab[0][ 1] = avg_rv40_qpel16_mc10_c;
c->avg_pixels_tab[0][ 2] = dsp->avg_h264_qpel_pixels_tab[0][2];
c->avg_pixels_tab[0][ 3] = avg_rv40_qpel16_mc30_c;
c->avg_pixels_tab[0][ 4] = avg_rv40_qpel16_mc01_c;
c->avg_pixels_tab[0][ 5] = avg_rv40_qpel16_mc11_c;
c->avg_pixels_tab[0][ 6] = avg_rv40_qpel16_mc21_c;
c->avg_pixels_tab[0][ 7] = avg_rv40_qpel16_mc31_c;
c->avg_pixels_tab[0][ 8] = dsp->avg_h264_qpel_pixels_tab[0][8];
c->avg_pixels_tab[0][ 9] = avg_rv40_qpel16_mc12_c;
c->avg_pixels_tab[0][10] = avg_rv40_qpel16_mc22_c;
c->avg_pixels_tab[0][11] = avg_rv40_qpel16_mc32_c;
c->avg_pixels_tab[0][12] = avg_rv40_qpel16_mc03_c;
c->avg_pixels_tab[0][13] = avg_rv40_qpel16_mc13_c;
c->avg_pixels_tab[0][14] = avg_rv40_qpel16_mc23_c;
c->avg_pixels_tab[0][15] = ff_avg_rv40_qpel16_mc33_c;
c->put_pixels_tab[1][ 0] = dsp->put_h264_qpel_pixels_tab[1][0];
c->put_pixels_tab[1][ 1] = put_rv40_qpel8_mc10_c;
c->put_pixels_tab[1][ 2] = dsp->put_h264_qpel_pixels_tab[1][2];
c->put_pixels_tab[1][ 3] = put_rv40_qpel8_mc30_c;
c->put_pixels_tab[1][ 4] = put_rv40_qpel8_mc01_c;
c->put_pixels_tab[1][ 5] = put_rv40_qpel8_mc11_c;
c->put_pixels_tab[1][ 6] = put_rv40_qpel8_mc21_c;
c->put_pixels_tab[1][ 7] = put_rv40_qpel8_mc31_c;
c->put_pixels_tab[1][ 8] = dsp->put_h264_qpel_pixels_tab[1][8];
c->put_pixels_tab[1][ 9] = put_rv40_qpel8_mc12_c;
c->put_pixels_tab[1][10] = put_rv40_qpel8_mc22_c;
c->put_pixels_tab[1][11] = put_rv40_qpel8_mc32_c;
c->put_pixels_tab[1][12] = put_rv40_qpel8_mc03_c;
c->put_pixels_tab[1][13] = put_rv40_qpel8_mc13_c;
c->put_pixels_tab[1][14] = put_rv40_qpel8_mc23_c;
c->put_pixels_tab[1][15] = ff_put_rv40_qpel8_mc33_c;
c->avg_pixels_tab[1][ 0] = dsp->avg_h264_qpel_pixels_tab[1][0];
c->avg_pixels_tab[1][ 1] = avg_rv40_qpel8_mc10_c;
c->avg_pixels_tab[1][ 2] = dsp->avg_h264_qpel_pixels_tab[1][2];
c->avg_pixels_tab[1][ 3] = avg_rv40_qpel8_mc30_c;
c->avg_pixels_tab[1][ 4] = avg_rv40_qpel8_mc01_c;
c->avg_pixels_tab[1][ 5] = avg_rv40_qpel8_mc11_c;
c->avg_pixels_tab[1][ 6] = avg_rv40_qpel8_mc21_c;
c->avg_pixels_tab[1][ 7] = avg_rv40_qpel8_mc31_c;
c->avg_pixels_tab[1][ 8] = dsp->avg_h264_qpel_pixels_tab[1][8];
c->avg_pixels_tab[1][ 9] = avg_rv40_qpel8_mc12_c;
c->avg_pixels_tab[1][10] = avg_rv40_qpel8_mc22_c;
c->avg_pixels_tab[1][11] = avg_rv40_qpel8_mc32_c;
c->avg_pixels_tab[1][12] = avg_rv40_qpel8_mc03_c;
c->avg_pixels_tab[1][13] = avg_rv40_qpel8_mc13_c;
c->avg_pixels_tab[1][14] = avg_rv40_qpel8_mc23_c;
c->avg_pixels_tab[1][15] = ff_avg_rv40_qpel8_mc33_c;
c->put_chroma_pixels_tab[0] = put_rv40_chroma_mc8_c;
c->put_chroma_pixels_tab[1] = put_rv40_chroma_mc4_c;
c->avg_chroma_pixels_tab[0] = avg_rv40_chroma_mc8_c;
c->avg_chroma_pixels_tab[1] = avg_rv40_chroma_mc4_c;
c->rv40_weight_pixels_tab[0] = rv40_weight_func_16;
c->rv40_weight_pixels_tab[1] = rv40_weight_func_8;
if (HAVE_MMX)
ff_rv40dsp_init_x86(c, dsp);
}